World Maritime Technology Conference,

WMTC2012, May 29-June 1, 2012, Saint-Petersburg, Russia

CALCULATION OF PROPULSION SHAFTING TRANSIENT

TORSIONAL VIBRATION INDUCED BY ICE IMPACTS
ON THE PROPELLER BLADES

Yuriy A. Batrak, Аnatoliy M. Serdjuchenko, Alexander I. Tarasenko

Intellectual Maritime Technologies
Krylova str., 19-g/12 Mykolaiv, Ukraine 54038

ABSTRACT

Calculation of transient torsional vibration induced by ice block impacts on the propeller blades is mandatory
since January 2011 when new Classification Societies Rules for the ships navigating in ice came into force. The
calculation complexity of shaft torsional vibration in this case consists not only in inevitable time-domain ap-
proach application but mostly in simulation of dynamic propulsion system response as a whole, taking into ac-
count shafting speed drop during impacts, diesel engine governor and turbocharger reaction for diesel engine
installations. Calculation time for the real propulsion trains grows considerably. As a way out Classification
Societies strongly advice to reduce actual mass-elastic systems to several mass using special technique. This
paper is concerned with an effective method for calculation of transient torsional vibration of the propulsion
system induced by ice block impacts that does not require mass-elastic system simplification. Special time-
domain integration of the linear matrix equations of propulsion shafting transient torsional vibration is pro-
posed. Calculation module based on this technique has been implemented in ShaftDesigner CAE package.

INTRODUCTION The time-domain approach for torsional vibration

Propeller of a ship navigating in ice is continually
subjected to ice impacts that, among other problems,
results in a severe shafting vibration. However the
international economy and exploitation of natural
resources of Northern areas requires elongating of
shipping period in Baltic Sea and in Arctic region.
That is the reason why numerous investigation pro-
jects were undertaken in the Northern countries last
decades. As a result of long-term efforts of interna-
tional research group new Rules for the ships navi-
gating in ice came into force in January 2011.
Among other requirements the updated Rules for the
ice and polar class ships require calculation of transi-
ent torsional vibration, caused by ice impacts on the
propeller blades.
Conventional propulsion shafting forced torsional
vibration calculations (TVC) have no computational
problems because owing to the harmonic excitation
law the frequency-domain approach can be easily
used.
In the transient vibration calculation time-domain
approach should be used because of the arbitrary ex-
citation law.
calculation of real propulsion trains is time consum-
ing considerably. As a way out the Classification So-
cieties, for example DNV [1], recommend to simplify
the conventional mass-elastic systems, used in forced
torsional vibration calculations, to several mass. Spe-
cial technique is to be applied for simplification of
the mass-elastic system. This technique is not trivial
and requires some experience from the user and
moreover it not be easily applied in the case of non-
linear elements.
The nature of propeller and ice block interaction re-
quires simulating of the propulsion system dynamic
response as a whole, taking into account shafting
speed drop, governor and turbocharger reaction for
diesel engine installations.
This paper is concerned with formulation of govern-
ing equations of propulsion shafting transient torsion-
al vibration caused by ice impacts on the propeller
blades and development of an effective method for
the time-domain integration of the equations that
gives the possibility to avoid the procedure of mass-
elastic system simplification and takes into account
propulsion system response during ice milling pro-
cess.

1
 World Maritime Technology Conference,
WMTC2012, May 29-June 1, 2012, Saint-Petersburg, Russia

1. GOVERNING EQUATIONS lumped mass. These arise due to the piston and
conrod center-of-mass positions changing with re-
Typical directly driven propulsion system equipped spect to the crankshaft axis and due to changing of
with controllable pitch propeller (CPP) (Fig. 1) con- propeller added inertia during pitch adjusting. See
sists of: articles [3,4], [5] for appropriate calculation formulas.
 diesel engine;
 propulsion shafting; Seven categories of the torques contribute to the
 CPP; sum  M k :
 oil distribution box, (ODB); – M kW , M kJ , M kP - weight, inertia and gas excita-
 speed governor;
tion torques are applied to the cylinder
 exhaust gas receiver; lumped masses;
 turbocharger; – M kH , M kI - hydrodynamic excitation and ice
 combustion air receiver;
impact torques are applied to the propeller
 load governor;
lumped mass;
 engine control system.
– M kD - absolute damping torque is applied
mainly to the cylinder and propeller lumped
Turbine
receiver
Turbocharger
masses;
– M kE - elastic torque produced by the stiffness
Load Speed
elements is applied to all lumped masses.
governor governor
Air cooler
Compressor
Sz nz
receiver TVC for open water operation condition when no ice
S
torque is applied usually performed in a frequency
domain as steady-state oscillations because shaft rota-
tion speed assumed to be constant as well as the rest
ODB governor
propulsion system parameters. Mean torque devel-
n
oped by the engine is in equilibrium with the mean
hydrodynamic torque, applied to the propeller.

Fig. 1: Propulsion system principal scheme

Mechanical part of the propulsion system in torsional

vibration calculation (TVC) is to be modeled as a
commonly known N-degree mass-elastic system: the
system consisting of lumped masses having specific
inertia, connected by the massless stiffness elements.

Motion of the lumped mass k of the system is de-

scribed by the equation [2]:

d  2k 

 kJ  J ( )     Mk , (1)
d 
k
2

where  k – rotation angle; k   k – angular veloci-

ty; J k – constant part of the lumped mass inertia;
J k () – variable part of the lumped mass inertia;
M k
– sum of the torques, applied to the lumped
mass.

According to [2] equation (1) can be rewritten as: Fig. 2: Ice torque graphs for 4-blades propeller

2k dJ k () In ice conditions the torques balance is disturbed by

 k J k  J k ()    M k   (2) ice impacts. The ice impact torques M kI in the Clas-
2 d
sification Rules are described as the sequences of
The variable inertia component J k () inherited half-sinus waves of blade passing frequency as well
as the double frequency. Ice torque graphs for 4-
mainly to the cylinder lumped mass and the CPP

2
 World Maritime Technology Conference,
WMTC2012, May 29-June 1, 2012, Saint-Petersburg, Russia

blades propeller at the constant propeller speed are 1 if n  nZ   ,

shown in the Fig 2. 
n  n
μD   Z if nZ  n  n Z   , (9)
Torque amplitude Qmax depends on ship class, propel-  
ler and hub diameters and propeller speed. 
0 if n  nZ ,

In the case of ice impacts propeller load changes ab- where  – speed control range, nZ – target speed:
ruptly that initiate variation of propulsion system pa-
rameters. First of all shaft actual speed n drops and
the speed governor increases the components of the  nZ for P governor,

fuel supply vector b sequentially to maintain required nZ   1
t
(10)
speed. Where, in spite of maximum fuel supply, the  nZ  T  (n  n) dτ for PI governor,

Z

i 0
required speed nz is not reached, load governor de-
creases the propeller pitch to decrease ship speed and
nZ – required engine speed.
use the engine energy to overcame ice resistance.
As a result of the fuel supply alteration exhaust gases
parameters will alter the combustion air parameters. Governor PID algorithm is not effective for low
In total these changes will alter indicating pressure speed installations and does not used here.
diagram and gas toque component at the cylinder
lumped masses [6]:
2. NUMERICAL TECHNIQUE
sin(α k  φk )
M kP  P(φk , PS* , TS , bk )  F  r  , (3) The choice of the numerical technique for the time-
sin α k
domain solution of the k+5 differential equations de-
where P(φk , PS* , TS , bk ) – cylinder pressure; PS* – air scribed above is a key factor for the successful prob-
lem solving. According to the fundamental publica-
receiver pressure; TS – air receiver temperature; bk –
tion [7] the majority of the numerical algorithms may
fuel supply vector component in cylinder k ; F – cyl-
be collected into the three families – multistage tech-
inder area; r – crank radius; α k – conrod angle. niques, multistep techniques and optimization tech-
niques.
For the propulsion shafting transient torsional vibra-
tion calculation induced by ice impacts the conven-
The disadvantage of the multistage techniques
tional TVC differential equation system must be ex-
(Runge-Cutta method, Newmark β-method) is that
panded with the differential equations for turbo-
some amount of the additional iterative calculations
charger rotor (4), turbine receiver (5)-(7) and com-
pressor receiver (8): for every temporal point t k 1 has to be carried out to
achieve required accuracy.
ntc  τtc = ftc ( P , P , T , Ltr , mtr , ntc , n),
* * *
(4)
S tr tr The well-known family of the “predictor-corrector”
mtr  τtr = f mtr ( P , P , T , Ltr , mtr , ntc , n, b ),
S
* *
tr
*
tr (5) algorithms (Adams-Multon method) belongs to the
multistep techniques. The disadvantage of the
Ltr  τtr = f L ( P , P , T , Ltr , mtr , ntc , n, b ),
S
* *
tr
*
tr (6) multistep techniques is that the algorithms could start
Ttr*  τtr = fT ( PS* , Ptr* , Ttr* , Ltr , mtr , ntc , n, b ), (7) only from some temporal point t s 1 , s  1 but not
mcr  τcr = f mcr ( P , mcr , ntc , n),
S
*
(8) from the initial point t0 , therefore some other algo-
rithm has to be used at the beginning for the calcula-
where ntc – turbocharger rotor speed, mtr – mass of
tions on the initial interval [t0 ; t s ] .
gas in turbine receiver, Ltr – gas amount in turbine
* *
receiver in moles, Ttr , Ptr – turbine receiver inhibited
flow temperature and pressure, mcr – mass of air in The third family of the optimization techniques is
compressor receiver; τ tc , τtr – time constants of based on the optimization procedures applied to the
compressor and turbine receivers; specified functional J (, , ) , connected to the con-
ftc , fmtr , f L, fT , ftcr – the right-hand sides functions of sidered problem. If { pn } are the set of free unknown
the equations [6].
parameters in the solution, then the conditions
J  pn  0 generate the resulting equations in the
Suitable fuel supply bk depends on the normalized
torque value μ D   M kP / M MCR defined by the speed algorithm. Different versions of the well-known least-
square algorithms are the good representatives of this
governor algorithm: family [9]. One of them is Kujawski&Gallager

3
 World Maritime Technology Conference,
WMTC2012, May 29-June 1, 2012, Saint-Petersburg, Russia

method [10], according to which the minimization

procedure is formulated as J  k 1  0 where func-

 t  Lt  
2 2  2  2
   Lk 1 k 1  Lk  k  Lk 1 k 1 , 
where Lk 1   12  1    and Lk  (1   ) are the
tional J has to be considered as a square of the total (2) (2) 2
errors of equations for the temporal inter-
Lagrangian polynomials of the second order for local
val [t k 1 ; t k 1 ] .
temporal coordinate   (t  tk ) t 1, 1 and
To solve the problem of transient torsional vibration k   tk  are the nodal values of the solution. The
problem the Kujawski&Gallager algorithm was spe- same polynomial approximation must be applied for
cially generalized [11] to be applied to the complete (2)
mechanical form of the governing equations with the vector of excitation forces T (t )  ( Lt  T ) in the
nonlinearities in matrix elements: right-hand side of the equations.

  C    T (t ) , (11)

where matrices ,C,  are considered as the equiva-

lent inertia, damping and rigidity matrices of the vi-
bration system respectively and T (t ) is a vector of
generalized excitation forces. For the transient and
nonlinear problems some elements of the above men-
tioned matrices would include dependences of the
matrix elements on time, displacement vector  and
velocity vector  . In this case loading vector may Fig. 3: Approximation of the solution on the temporal
interval [t k 1 ; t k 1 ] ; I – known portion of the solution,
have the same structure too T (t , , ) .
II – unknown portion of the solution.
The mechanical form of the equation makes the gen-
eralised Kujawski&Gallager method most suitable for
Approximation generates some errors in the solution
use together with the finite element method of pro-
or so called “residual forces” in the matrix equations
pulsion shafting modelling [8].
(11)
To evaluate the above mentioned techniques four      
  t    Lt   C Lt    Lt   Lt T
2 2 2
  2
  0
typical algorithms – fourth-order Runge-Cutta algo- Then, at the second stage of the algorithm [8,11], the
rithm, four-step Adams-Multon algorithm, Newmark functional related to the problem can be formulated as
β-method and generalized version of Ku- the following:
jawski&Gallager algorithm have been chosen to solve
the following differential equations: 1
1) linear dumping oscillator, T
 
J R   W   c   c  d ,  (14)
2) nonlinear Duffing’s oscillator and 1
3) Van-der-Pol’s oscillator with nonlinear damping.
Calculations were carried both for free oscillations where W    is some weighting function on the con-
and for harmonically exited oscillations. sidered temporal interval having the properties
Calculation results brought us to the following con-
1
  
1W  d  1, W   W  , W   0   and
clusion: every of chosen algorithm works correctly
for some of the equations and indicates amplitude
       
c  t    c Lt   Cc Lt  w   c Lt  w  Lt  wT
2 2 2 2

instability or phase shifting for the rest of the equa- is the vector of residual forces. This vector includes
tions. Generalized Kujawski&Gallager optimization additional weighting factors w  diag   1 ;  0 ;1 and
algorithm in general showed better characteristics in estimating of the matrixes ,C,  in some specified
the comparison to other algorithms.
point of collocation t  tc  [tk 1 ; t k 1 ] (in the case of
Hereinafter generalized Kujawski&Gallager algo- nonlinear or time-depending elements of matrixes in
rithm [8,11] is discussed in more details. the equations). Weighting function W  t  , point of
At the initial stage of the algorithm, for the constant collocation t c and weighting factors w have to be
temporal step t  const the second-order polyno- chosen on the next stages of the algorithm.
mial approximation of the solution is applied on the
temporal interval [t k 1 ; t k 1 ] (see Fig. 3):

4
 World Maritime Technology Conference,
WMTC2012, May 29-June 1, 2012, Saint-Petersburg, Russia

At the third stage, the algorithm considers k 1 , k the first temporal point t1 and stability conditions at
as known values; for the unknown value  k 1 we the infinity tn   . Application of these conditions
would formulate minimization procedure as the fol- generates finally the following results for the weight-
lowing: ing factors k j , j  1, 2, 3,4 [11]: relations between the
factors
J R
0, (15)
 k 1
 
k1  12 , k3  12   1   0 ;   
 (19)
which finally results in the numerical algorithm:   1
k 2   1  1  k 4  24 1   2 ,
  
 k 1  2 A
1

B k 1    2 A
1

D  k 1 
(16)
accuracy and stability conditions for two free tuning
factors , k 4
t
2  A1E Tk 1  Tk 1   2 A1  GTk  HTk 1  ,


where k  1, 2, 3, .... and for the first temporal point t1

5
   7

12  1   3  k 4 1   0  180 1   4  0,
  

1  4  
it modifies to the equation:  
1  2   16   k 4     12
2
*  
 *   (20)

 
1  1  (17) 
1
 1

1

 k 4  4   12    12   0,
6
 
 B0  t  A  2 D  0   2 ET1  GT0 
1  A  D
*
 
where  j , j  0, , 4 are some correction factors re-
where 0 , 0 are the initial values of the problem lated to the damping effects in the model problem and
solution. 2 2
  t (0  v ) .
2 2
*
The following expressions for the matrices A, B, D,
E, G, H are used:

 T 2 T

A   c  c  t  c C c  t k1  c  c   c  c  2C c C c  
T T T
 
3 T
 T
 12 t k1  c C c  2C c  c  4 T
t k 2  c  c ,  

T
B   c  c  t
2 T
 T T
k3  c  c  k1 c  c  2 k1C c C c    (18)

4 T
+ t k 4  c  c ,  

 T 3 T
 T
D  t  c C c  12 t k1  c C c  2  c C c  , 

 T T 2 T
E  k1 c  tk1C c  t k 2  c  , G  k Tc  t 2k Tc  ,
3 4 

H  tk C c  .
T
1 

The algorithm includes four unknown weighting fac-

tors k j , j  1, 2, 3,4 which are some integrals from
weighting function W    and approximation poly-
Fig. 4: Stability diagram for tuning factors , k 4 :
(2) (2)
nomial Lk , Lk 1 . These factors must be estimated by 1, , 5 – the upper boundaries of conditional stability
the application of the algorithm to some model prob- for values   1.0; 1.25; 1.60; 1.80; 5.0 respectively;
lem, associated with the main problem.
6 – eighth-order accuracy condition for testing prob-
2 lem: k4  125   7  0 ; /////// – domain of the abso-
Linear damping oscillator u  2u  0 u  0 , t  0 180
with the initial conditions u (0)  1, u (0)  0 can be lute stability; the specific points on the diagram: A –
1 , k  1 ); B
(   24 – (   0 , k 4  14 ) ; C –
used as a model problem for transient torsional vibra- 4 48
tion calculation. 1 1 ).
(   4 , k 4   24
Applying the algorithm to the model problem we
would take into account the accuracy conditions for The accuracy and stability conditions in equation (20)
can be displayed on the algorithm stability diagram

5
 World Maritime Technology Conference,
WMTC2012, May 29-June 1, 2012, Saint-Petersburg, Russia

for the testing problem, Fig. 4. This diagram may be For both elements the maximal torsional vibration
used to specify two free weighting factors , k 4 in stress 8.5-16.8 MPa arise within speed interval 60-75
rpm. Vibration stress component of order 6 prevail in
the solution of the actual problem. For example in the the synthesis stress.
1 , k  1 can be
first approach the values   24 4 48
adopted in the calculations.

3. CALCULATION SAMPLE

Calculation algorithm is intended to be implemented

in ShaftDesigner software [12] as a separate module
for transient ice impact torsional vibration analysis.
Currently this module does not take into account shaft
speed drop due to the ice impacts. Some calculation
results for IACS polar class PC1 vessel, equipped
with low speed installation are discussed below.

Fig. 7: Steady vibration stress in the throw 6

The transient vibration stress during ice milling is

much higher: 165-348 MPa for the propeller shaft
(Fig. 8) and 51-96 MPa for the throw 6 (Fig. 9).

Fig. 5: Mass-elastic system for low speed installation

Main characteristics of the propulsion installation are

as the following:

Engine stroke 2
Cylinder number 6
MCR 5000 kWt
Rated speed 109 rpm
Propeller type FPP
Blade number 4
Propeller shaft diameters 410/130

Steady vibration stress in the propeller shaft and in

the throw 6 (aftermost cylinder marked as lumped
mass 8) for open water operation condition are shown Fig. 8: Transient vibration stress in the propeller
in the Fig. 6 and Fig 7. shaft

Fig. 6: Steady vibration stress in the propeller shaft Fig. 9: Transient vibration stress in the throw 6

6
 World Maritime Technology Conference,
WMTC2012, May 29-June 1, 2012, Saint-Petersburg, Russia

Maximum torsional stress due to the ice impacts, in In the Fig. 10-13 transient torsional vibration graphs
contradistinction to open water condition, arise within are shown. As can be seen ice impact torsional stress
the speed interval 95-100 rpm. It is exactly the same decay very fast after ice torque becomes equal to
interval where the first-blade order resonance is lo- zero.
cated (see order 4 curve in the Fig. 6). Classification
societies recommend analyse torsional vibration due
to the ice impacts at this location first of all.

Fig. 13: Transient vibration stress in the throw 6 at

100 rpm

Fig. 10: Transient vibration propeller shaft stress at

68 rpm CONCLUSIONS

Proposed algorithm was programmed and included to

ShaftDesigner CAE package as a separate module for
transient torsional vibration calculation. Applications
of this module demonstrate quite acceptable calcula-
tion time for full mass-elastic system. Hence no sim-
plification of mass-elastic system is required to calcu-
late transient torsional vibration in ice operating con-
ditions.

As a first stage of algorithm implementation shaft

speed drop was not taken into account. Such option
gives the possibility for the fast evaluation of tor-
sional vibration stress because it does not require of
whole propulsion shafting system modelling. It is a
Fig. 11: Transient vibration propeller shaft stress at decisive argument at the early stage of propulsion
100 rpm shafting design. In addition using this option we are
on the safe side. It means that if the calculated pa-
rameters satisfy the acceptance criteria, more detailed
calculation may not be required at all.

The option for the whole propulsion system model-

ling is under testing now. It will be useful when the
fast modelling results do not satisfy acceptance crite-
ria or real propulsion system parameters are of inter-
est. In the last case the realistic not statutory ice
torques are desirable to have as an input.

ACKNOWLEGMENTS

We express special thanks to Stanislav Avanesov and

Erik Werner of Germanischer Lloyd for their contri-
Fig. 12: Transient vibration stress in the throw 6 at butions and feedback.
68 rpm

7

WMTC2012, May 29-June 1, 2012, Saint-Petersburg, Russia
REFERENCES
1. Ice strengthening of propulsion machinery, 2010
DNV Classification Notes No 51.1, September, 63 pp.
2. Frolov К.V., Popov S.А., Мusatov А.К., and oth-
ers., 1987, Mechanisms and Machines Theory –
Мoscow: «Vysshaya Shkola», 496 pp. (in Russian)
3. Hesterman D.C., Stone B.J., 1994, A Systems Ap-
proach to the Torsonal Vibration of Multi-Cylinder
Reciprocating Engines and Pumps. Proceedings of
the IMechE, Part C: Journal of Mechanical Engi-
neering Science 208 (6), pp. 395-408.
4. Guzzomi A.L., 2007, Torsional Vibration of
Powertrains: An investigation of Some Common As-
sumptions. PhD thesis, University of Aestern Austral-
ia. 136 pp.
5. MacPherson D.M., Puleo V.R., Packard M.B.,
2007, Estimation of entrained water added mass
properties for vibration analysis. SNAME New Eng-
land Section, June.
6. Тarasenko А.I., 2008, Non-linear Dynamic Model
of Low Speed Marine Engine. Vestnik
Dvigatelestroyeniya: Zaporzhzhia: ”Мotor
Sich”, 3 (20), pp. 202–205. (in Russian)
World Maritime Technology Conference,
7. Hull G., Watt J.M., 1976, Modern Numerical
Methods for Ordinary Differential Equations, Clar-
endon Press. Oxford, 312 pp.
8. Brebbia C.A., Walker S., 1979, Dynamic Analysis
of Offshore Structures. London-Boston, 230 pp.
9. Eason E.D., 1984, A review of list-squares meth-
ods for solving partial differential equations. Trans.
Soc. Math. Eng. 1, pp. 1021-1046.
10. Kujawski J., Gallager R., 1985, An efficient
higher order algorithm for nonlinear transient dy-
namic analysis. Proceed. NUMETA – 85 Conference
Swansea, pp. 51-58.
11. Serdjuchenko A.N. Dynamics of waves and ships
in sea conditions including nonlinear effects. Applied
Hydromechanics, Vol. 72. Kiev, pp. 112-135 (in Rus-
sian)
12. Y. A. Batrak., 2009, New CAE Package for Pro-
pulsion Train Calculations, ICCAS papers, Septem-
ber Shanghai, v II, pp.187-192.